Case study: Mediahuis

Mediahuis, the largest media company in Belgium, owns four of Belgium’s most-read newspapers and caters to millions of readers daily. They launched an optimization program with the specific goal of making change based on data rather than speculation, and dedicated a team specifically to company-wide experimentation. "Right now we're testing for learning and to convince the right people—we want to get rid of testing by gut feeling and to become more data-driven," says Yves Van Dooren, Manager of the Digital Traffic Center. As a result, Mediahuis has the resources at their disposal to run quick experiments on changes before launching them to their entire audience, greatly reducing risk and development time.

The team’s first task was to optimize the subscription experience for users. As a significant income stream, improving the subscription process was key to building growth for the company. The team hypothesized they would see an improvement in click-through rates if they improved the outdated designs on the subscription pages. However, when they ran the experiments on the redesign, they quickly realized that design changes had no impact on the results. In fact, the changes they made actually had a negative impact! Running these experiments enabled Mediahuis’s optimization team to quickly adjust their strategy. Based on a suggestion from Optimizely, they experimented on the placement of new copy on their pages. When they analyzed their results, they found an uplift in click-through of more than 50% compared to all previous experiments.

In running quick experiments on their proposed changes, Mediahuis was able to avoid losses from a redesign that could have taken a great deal of effort to build and implement and found surprising success though something as seemingly simple as an adjustment to their copy. When they took the time to hypothesize and analyze their results, they actually saved time in the long run and learned a great deal about their audience and the value of experimentation overall.